During the COVID-19 pandemic, many institutions such as universities and workplaces implemented testing regimens with every member of some population tested longitudinally, and those testing positive isolated for some time. Although the primary purpose of such regimens was to suppress disease spread by identifying and isolating infectious individuals, testing results were often also used to obtain prevalence and incidence estimates. Such estimates are helpful in risk assessment and institutional planning and various estimation procedures have been implemented, ranging from simple test-positive rates to complex dynamical modeling. Unfortunately, the popular test-positive rate is a biased estimator of prevalence under many seemingly innocuous longitudinal testing regimens with isolation. We illustrate how such bias arises and identify conditions under which the test-positive rate is unbiased. Further, we identify weaker conditions under which prevalence is identifiable and propose a new estimator of prevalence under longitudinal testing. We evaluate the proposed estimation procedure via simulation study and illustrate its use on a dataset derived by anonymizing testing data from The Ohio State University.

翻译：在COVID-19疫情期间，许多机构（如大学和工作场所）实施了纵向检测方案，对特定人群的每位成员进行定期检测，检测阳性者需隔离一段时间。尽管这些方案的主要目的是通过识别和隔离感染者来抑制疾病传播，但检测结果也常被用于获取患病率和发病率估计值。此类估计值有助于风险评估和机构规划，目前已有多种估计方法被实施，从简单的检测阳性率到复杂的动力学建模。然而，在许多看似无害的纵向检测与隔离方案下，流行的检测阳性率作为患病率估计量存在偏差。我们阐释了此类偏差的成因，并明确了检测阳性率为无偏估计量的条件。进一步地，我们识别了患病率可识别的较弱条件，并提出了纵向检测下患病率的新估计量。通过模拟研究评估了所提估计方法，并在俄亥俄州立大学匿名化检测数据上展示了其应用。